61
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
Stress Testing the Canadian Banking System:
A System-Wide Approach
Kartik Anand, Guillaume Bédard-Pagé and Virginie Traclet
ƒƒ Stress testing is an important tool for evaluating risks to
the financial system. The models used to conduct these
tests are evolving to include more realistic features.
ƒƒ The 2007–09 financial crisis demonstrated that, in
addition to solvency risk, liquidity risk and spillover
effects can generate losses for banks during times of
stress. The Bank of Canada has developed an innovative stress-testing model—the MacroFinancial Risk
Assessment Framework (MFRAF)—which captures the
various sources of risk (solvency, liquidity and spillover
effects) that banks face.
ƒƒ We apply MFRAF to the stress-testing scenario used
in the 2013 Canada Financial Sector Assessment
Program led by the International Monetary Fund. We
show that the aggregate capital position of Canadian
banks is 20 per cent lower when liquidity and spillover
risks are added to solvency risk. Nevertheless, the
results still confirm the overall strength of the Canadian
banking system.
Introduction
Over the past few years, financial sector authorities and
financial institutions around the world have increased
their use of stress testing to examine risks to the financial
system. Stress testing assesses the impact of various
potential risks to financial institutions and illustrates the
channels through which these risks would be transmitted.
While most stress-testing models focus on solvency risk
(the risk of losses stemming from the failure of borrowers
to repay loans or meet contractual obligations), the
2007–09 financial crisis showed that, in times of stress,
liquidity risk and network spillover effects associated with
interconnections among banks can also be significant.
The Bank of Canada has developed an innovative stresstesting model, the MacroFinancial Risk Assessment
Framework (MFRAF), which realistically captures the
various sources of risk for banks—solvency risk, liquidity
risk and spillover effects.
In 2013, Canada participated in a Financial Sector
Assessment Program (FSAP), a comprehensive, in-depth
analysis of the country’s financial sector conducted by
the International Monetary Fund (IMF) that included a
stress-testing exercise to gauge the resilience of financial
institutions to severe macrofinancial stress.1 The FSAP
stress scenario embodied the realization of two key risks
to the Canadian financial system that had been identified in previous issues of the Financial System Review:
(i) weaknesses in euro-area banks and sovereigns, and
(ii) imbalances in Canadian household finances and the
housing market. Several stress-testing approaches and
models, including MFRAF, were used to estimate the
impact of these risks on the Canadian banking system
should they be realized. Overall, the results confirm the
strength of the Canadian banking system as a whole,
and the IMF views the resulting capital shortfall as
manageable.
The results obtained with MFRAF show that, when
liquidity risk and spillover effects are considered in
addition to solvency risk, the aggregate capital position
of banks declines by an additional 20 per cent.2 This
finding highlights the importance of a comprehensive
approach to stress testing. As well, it creates an incentive for the Bank of Canada to further enhance MFRAF
to improve its understanding of the potential effects
of a severe stress scenario on the Canadian banking
system.
The following three sections: (i) define stress testing
and how it is used to assess risks; (ii) describe
MFRAF; and (iii) present the results obtained with
MFRAF in the context of the 2013 FSAP. The final section concludes with remarks on the direction of future
research.
1 For the conclusions of the 2013 Canada FSAP, see IMF (2014).
2 The IMF views MFRAF as being “at the frontiers of systemic risk stress
testing.” See IMF (2014), page 54.
62
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
Table 1: Stress testing: Comparing models and approaches
Key features
Bottom-up stress test
Top-down stress test
Run by individual banks
Run by authorities
Testing for
Solvency risk
Strengths
ƒƒ Bank models capture a large
amount of detailed data on
their portfolios and exposures,
which provide information on
the specific drivers of their
stress-testing results.
ƒƒ Banks have different business models and exposures
to risks: their stress-testing
models capture these idiosyncrasies.
Limitations
ƒƒ Interactions with other banks
during periods of stress and
related network effects are not
accounted for.
ƒƒ Liquidity risk is not explicitly
captured (beyond the effects
of rising funding costs in times
of stress).
Solvency risk
ƒƒ Authorities use a common model to generate results for different banks, enabling
comparison of the results across banks to
better understand their respective vulnerabilities to specific shocks.
ƒƒ Tests are usually applied on a bank-by-bank
basis, but results can be aggregated to
determine the “typical” impact of specific
stress scenarios on the banking sector.
MFRAF: solvency risk, liquidity
risk and spillover effects
ƒƒ Sources of risks affecting banks
are taken into account in a consistent manner.
ƒƒ Tests are applied on a bankby-bank basis, but interactions
between banks are captured.
ƒƒ The framework can be used as a
pure top-down model or as a “hybrid,” with inputs from a bottomup stress test.
ƒƒ Tests provide fewer details regarding the
drivers of results than in a bottom-up stress
test.
ƒƒ Tests provide fewer details on
drivers of results than bottom-up
stress tests.
ƒƒ Tests use simple models based on observed historical relationships between key
macrofinancial variables and banks’ indicators, making it more difficult to capture the
idiosyncrasies of individual banks.
ƒƒ This approach requires large
amounts of detailed balancesheet data.
ƒƒ Liquidity risk is not explicitly captured
(beyond the effects of rising funding costs in
times of stress).
ƒƒ Interactions between banks are not explicitly
taken into account; hence, there are no
spillover effects.
Source: Bank of Canada
Stress Testing: Definition, Uses and
Components
Stress testing is a tool used by banks for purposes of
internal risk management and by authorities to quantify
the impacts that large but plausible negative shocks
could have on the capital positions of banks (BCBS 2009).
Stress tests do not take into account corrective management actions such as raising additional capital and
implementing cost-cutting measures, which banks would
typically take if such conditions were to materialize.3 In this
sense, stress tests evaluate extreme outcomes.
There are two main approaches to conducting stress
tests. In bottom-up stress tests, individual banks use
their internal models. In top-down stress tests, regulatory authorities apply their own models. The chief
advantage of a bottom-up stress test is that, since
banks’ internal models capture each institution’s idiosyncrasies, it is possible to better understand the specific
3 Management responses to financial stresses are part of the recovery plans
that systemically important financial institutions are expected to design as
part of the G-20 regulatory reform agenda. In Canada, the Big Six banks
have been designated as domestic systemically important banks (D-SIBs)
by the Office of the Superintendent of Financial Institutions and, consequently, are required to develop recovery and resolution plans. D-SIBs
are required to hold a 1 per cent common equity surcharge starting in
2016 (i.e., they will be required to meet an 8 per cent CET1 ratio). See OSFI
(2013a).
drivers of the results for individual banks. In contrast,
the main advantage of a top-down stress test is that, by
using a common model for different banks, authorities
can compare the results across banks to obtain insights
regarding their respective vulnerabilities to the same
shocks. However, weaknesses are also evident in both
approaches. It can be more difficult, for example, to
take into account the interactions between banks in
a bottom-up stress test, while top-down tests tend to
capture the characteristics of banks in less detail.
Table 1 summarizes the key strengths and limitations of
the various stress-testing approaches.
Stress testing is being used more frequently by authorities around the world, but in different ways. In some
jurisdictions, the focus is on the stress-testing results for
individual banks. For example:
ƒƒ In the United States, the Federal Reserve evaluates
plans by large banks to make capital distributions and
approves these plans only for institutions that demonstrate sufficient financial strength under a severe
stress scenario.4
4 For example, in 2014, the Federal Reserve objected to the plans of five of
the 30 participating banks. For details, see http://www.federalreserve.gov/
newsevents/press/bcreg/20140326a.htm.
63
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
ƒƒ In Europe, authorities have used stress tests to
evaluate the resilience of individual European banks
and to assess their recapitalization needs under
stressed conditions. Before assuming its supervisory
role in November 2014, the European Central Bank
(ECB) will conduct and publish a stress test as part
of its comprehensive assessment in order to rebuild
investor confidence in the European banking sector.
In other jurisdictions, e.g., Sweden and Norway,
authorities use stress tests to better understand how
the banking sector would be affected by adverse
macroeconomic developments (Sveriges Riksbank 2012;
Norges Bank 2013). In Canada, stress testing is part
of the tool kit used to assess risks for individual banks
and for the banking sector as a whole. The Office of the
Superintendent of Financial Institutions (OSFI) promotes
internal stress testing as an important tool for banks to
use in making decisions related to business strategy,
risk management and capital management.5 In this context, OSFI reviews institutions’ stress-testing programs
as part of its supervisory review process and its review
of the internal capital-adequacy assessment process
for deposit-taking institutions. Moreover, OSFI and the
Bank of Canada conduct an annual joint exercise to
stress test the major Canadian banks to identify systemwide vulnerabilities that could materialize under adverse
macrofinancial conditions, and use the results to inform
assessments of risk for the financial system as a whole.6
This joint exercise is a bottom-up stress test: although
the stress scenario and detailed instructions for applying
the stress test are designed by the authorities, the banks
use their internal models to calculate the impact of the
stress scenario on their capital positions. The authorities
analyze and compare the results provided by individual
banks to determine the effects of the stress scenario
on the entire banking sector, with a particular focus
on understanding key drivers and the channels for the
transmission of shocks.
Most stress-testing models, whether top-down or
bottom-up, focus primarily on solvency risk.7 However,
as the financial crisis demonstrated, banks can be
significantly affected by two other sources of risk during
periods of stress: liquidity risk and spillover effects.
Liquidity risk results from the combination of fundingliquidity risk (the risk that individual banks are unable to
roll over existing funding or to obtain new funding) and
5 For details, see OSFI (2009).
6 This regular stress-testing exercise was implemented following a recommendation by the IMF during Canada’s 2007 FSAP.
7 Nevertheless, a range of market risks (including funding liquidity) that are
consistent with the stress scenario tend to be indirectly captured by these
models. For example, when the stress scenario incorporates conditions of
scarce funding liquidity and declining asset prices, the former would lead
to rising funding costs and interest expenses for banks and the latter to
mark-to-market losses on securities available for sale. These effects would
ultimately influence banks’ capital positions.
market liquidity conditions (the conditions under which
banks can sell and repurchase, or sell outright, assets
in financial markets to meet their funding needs). During
the financial crisis, interactions between funding liquidity
and market liquidity created liquidity spirals, which
particularly impacted institutions that relied heavily on
wholesale funding and held highly illiquid assets (e.g.,
Northern Rock and Bear Stearns), ultimately affecting
global financial stability.8 Network spillover effects occur
when a bank is unable to fulfill its obligations to other
banks, creating counterparty credit losses for those
banks (e.g., the banks exposed to Lehman Brothers
when it defaulted in September 2008).
In addition to accounting for solvency risk, MFRAF also
incorporates liquidity risk and network spillover effects.9
MFRAF: Model Description
MFRAF consists of three distinct, but interdependent,
modules that account for the three different risks
that banks face.10 Figure 1 shows how these risks
could materialize over a one-year horizon following
a risk event—for example, a severe macroeconomic
shock—and how they contribute to an aggregate decline
in the capital positions of banks. This decline is mea­
sured by determining the banks’ common equity Tier 1
(CET1) capital ratio, as follows.11 First, banks’ balance
sheets are affected by credit losses due to corporate
and household defaults six months into the first year
(the interim date) and again at the end of the first year.
Second, if investors have concerns about a bank’s
future solvency and/or its liquidity position, liquidity risk
materializes at the interim date, potentially generating
additional losses. Finally, at the end of the period, some
banks may be unable to repay their interbank counterparties, given the solvency and/or liquidity losses that
they have incurred, which leads to network spillover
effects. MFRAF considers each bank individually but
takes into account the interactions between banks
through both liquidity and interbank exposures.
Overall, the three risks contribute to a decline in the
capital positions of banks. By decomposing the decline
in CET1 ratios into their solvency, liquidity and network
8 See, among others, Brunnermeier (2009) and Brunnermeier and Pedersen
(2009). The Basel III liquidity framework was introduced to address the
failures in liquidity-risk management that were exposed by the financial
crisis. See Gomes and Wilkins (2013).
9 A number of other central banks (e.g., the Bank of England, the ECB and
the Bank of Korea) are also developing stress-testing models that capture
risks beyond solvency, although their methodologies differ.
10 See Appendix A for a more detailed description of the model and its
calibration.
11 The CET1 ratio is equal to common equity (the highest-quality capital)
divided by total risk-weighted assets.
64
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
Figure 1: MFRAF: A modular approach to systemic risk
Stress
scenario
Start of year (t0)
CET10
Interim date:
6 months (t1)
Solvency risk
Liquidity risk
CET11
Note: CET1 is the common equity Tier 1 capital ratio.
End of year (t2)
Solvency risk
Network spillover effects
Aggregate losses
CET12
Source: Bank of Canada
components, MFRAF contributes to a better understanding of the various determinants of risk for banks and
the channels through which shocks would propagate.
Solvency-risk module
In MFRAF’s solvency-risk module, banks’ balance
sheets are affected by credit losses that result from the
failure of non-bank borrowers to repay their loans or to
meet their contractual obligations under stress. For each
bank, we derive a distribution of expected annual credit
losses that takes into account the historical correlations
of defaults across sectors, together with the loss-givendefault rates and exposures at default to the different
sectors to which banks lend.12
Liquidity-risk module
In MFRAF, banks can be affected by liquidity risk, either
directly, through the funding decisions of their creditors,
or indirectly, through information contagion. Both of
these dynamics were observed during the financial
crisis. Liquidity risk can materialize endogenously as a
result of solvency risk and the liquidity characteristics of
banks (reliance on unstable funding and/or low holdings
of liquid assets). Following the realization of credit losses
at the interim (six-month) date (Figure 1), the creditors of
each bank must decide whether or not to roll over their
funding to the bank (i.e., whether to “run”). This decision
depends on two elements: (i) creditors’ concerns over
12 The sectors include households (uninsured residential mortgages, home
equity lines of credit and consumer loans), businesses (manufacturing, construction, accommodations, commercial real estate, agriculture, wholesale,
financial institutions and small business loans) and governments.
the future solvency of the bank (which depends on the
severity of the losses incurred by the end of the year and
the bank’s starting capital position) and (ii) the bank’s
liquidity characteristics.
Creditors assess a bank’s solvency relative to a certain
threshold (typically a supervisory threshold).13 In such
assessments, they compare the value of the bank’s
liquid and illiquid assets with its liabilities that are
susceptible to a run at the interim date. If the value of
the liquid and illiquid assets is greater than the stock of
liabilities susceptible to a run, the creditor judges that
the bank has more than enough liquidity to meet the
demands of all its creditors, and funding will be rolled
over. If the reverse is true, there is a positive probability
of a run; this probability is determined as the outcome of
a coordination game.14 When liquidity risk materializes,
banks experience additional losses.15
13 For the FSAP, we assumed that creditors have concerns about solvency
when a bank’s future CET1 ratio falls below OSFI’s supervisory threshold
of 7 per cent. It is important to note that falling under 7 per cent is not
equivalent to failure: the threshold for the Basel III regulatory CET1 ratio is
4.5 per cent.
14 A coordination game is a situation in which agents realize gains when they
all take the same action but make their decisions independently and are
uncertain about the actions of other agents. In this coordination game,
creditors compare the expected returns from running on the bank versus
rolling over their claims. For an individual creditor, the return from rolling
over its claims depends on the share of other creditors that also roll over
their claims. In contrast, a creditor that decides to run obtains a fixed return
(from investing instead in a risk-free asset).
15 For the FSAP, the liquidity losses were calibrated at 2.25 per cent of riskweighted assets.
65
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
Liquidity risk can also materialize because of information contagion, i.e., the risk that creditors will run on a
bank with a sound balance sheet after observing the
CET1 ratio of one or more other banks decline below
7 per cent.16 In this context, a bank’s creditors update
their beliefs regarding market liquidity conditions. In
some instances, creditors may become more pessimistic, which leads them to have a less favourable view
of the liquidity characteristics of their own bank and
influences their decision on whether to extend funding.
If the new-found pessimism is widespread, it may result
in contagious runs such as those observed during the
financial crisis.
The endogenous materialization of liquidity risk resulting
from the interactions between solvency, funding and
market-liquidity risk is a feature of MFRAF that sets it
apart from other stress-testing models.
Network spillover effects
Following the realization of credit and liquidity losses,
some banks may be unable to repay their full obligations to other banks. We consider interbank exposures
to be subordinate to other debt, i.e., banks first settle
other debt obligations before turning to their interbank
counterparties.17
Application of MFRAF in the 2013 FSAP
FSAP stress scenario
The stress scenario used in the 2013 FSAP includes the
materialization of the key risks identified in the Financial
System Review, which could arise from two areas:
(i) weaknesses in euro-area banks and sovereigns, and
(ii) imbalances in Canadian household finances and the
housing market.18 The stress scenario covered the fiveyear period from 2013 to 2017.
The stress scenario begins with a disorderly default in a
peripheral euro-area country, which results in a severe
and persistent economic recession and a renewed
banking crisis in the euro area. This leads to a general
retrenchment from risk in the global financial system and
significant disruptions in global bank funding markets,
causing important adverse confidence and wealth
effects and a weakening global economy. The Canadian
economy faces financial headwinds, a large negative
foreign demand shock, falling commodity prices, rising
uncertainty, and unfavourable effects on confidence
and wealth that affect both businesses and households, leading to a sharp decline in domestic demand.
Business investment and consumer spending decrease
significantly. As Canadian households face negative
wealth shocks, tighter lending standards, deteriorating
employment prospects and heightened uncertainty, they
significantly reduce their expenditures on consumption
and residential investment. In this environment, house
prices decline markedly. Consequently, Canada faces a
severe and persistent recession.19 As shown in Table 2,
the recession in the FSAP stress scenario is much more
severe than any recession experienced by Canada over
the past three decades.
Stress-test results
Overview
Four approaches were used in the FSAP to assess the
impact of this stress scenario on Canadian banks: (i) a
bottom-up solvency stress test conducted by the Big
Six Canadian banks; (ii) a top-down solvency stress
test conducted by OSFI; (iii) a top-down solvency stress
test conducted by the IMF; and (iv) MFRAF, which was
used as a “hybrid” model to complement the banks’
bottom-up solvency stress test by capturing the impact
of liquidity risk and network spillover effects.20, 21 In all
four approaches, banks were not allowed to include any
Table 2: Key macroeconomic variables in the Financial Stability Assessment Program stress scenario
Macroeconomic variables
Real GDP contraction (peak to trough, per cent)
Duration of recession (number of consecutive
quarters of negative growth)
Peak increase in unemployment rate
(percentage points)
House price correction (peak to trough, per cent)
2013 FSAP
2007–09 recession
1990s recession
1980s recession
-5.9
-4.2
-3.4
-5.1
9
3
4
6
5.9
2.4
4.1
5.8
-33.0
-7.6
-10.1
-4.2
Source: Bank of Canada
16 Information contagion is a recent innovation in MFRAF. Its inclusion enhances
the model’s ability to capture an important transmission mechanism observed
during the crisis. For details, see Anand, Gauthier and Souissi (2014).
17 To clear the interbank network, MFRAF uses the algorithm of Eisenberg and
Noe (2001), in which banks repay their interbank counterparties a sum that is
proportional to the amounts originally due, causing counterparty credit losses.
18 Note that this scenario was generated in early 2013 and was based on the
level of risks observed at that time. Since then, those risks have declined.
19 In this scenario, there is no liquidity injection by the central bank or extraordinary monetary policy stimulus.
20 For more information on the features of the various models and detailed
results, see IMF (2014).
21 In practice, using MFRAF as a hybrid to augment the bottom-up stress test
means that various outputs provided by the banks in the bottom-up stress
test are used as inputs for MFRAF.
File information
(for internal use only):
66
chart 1 Stress
-- EN.inddTesting the Canadian Banking System: A System-Wide Approach
Bank03:21:35
of Canada
• Financial
Last output:
PM; Mar
14, 2012 System Review • June 2014
The value added by MFRAF
Chart 1: Results of the IMF Financial Sector Assessment
Program stress test
Aggregate CET1 ratio for the Big Six banks
%
9.0
OSFI target CET1 ratio
(with D-SIB 1% surcharge)
8.5
8.0
7.5
OSFI target
CET1 ratio
7.0
6.5
6.0
5.5
Recovery
2013
2014
2015
Stress horizon
Big Six banks’ bottom-up stress test
OSFI top-down stress test
2016
2017
5.0
4.5
IMF top-down stress test
MFRAF
Note: CET1 is the common equity Tier 1 capital ratio.
Sources: Bank of Canada, International Monetary Fund, Office of the
Superintendent of Financial Institutions and major banks
corrective management actions, except for the Basel III
capital conservation buffer, which dictates restrictions
on capital distribution, depending on the level of the
CET1 ratio.22, 23
Chart 1 shows the dynamics of the aggregate CET1
ratio for the Big Six banks over the stress horizon under
each of the four approaches. Although the bottom-up
stress test and the OSFI and IMF top-down stress tests
capture the impact of solvency risk, there are some
differences in the results, which primarily reflect differences in modelling. Overall, under this stress scenario,
solvency risk results in a decline of 170 to 250 basis
points (from 8.33 per cent) in the aggregate CET1 ratio of
banks. Although this is a large decline, it is not surprising,
given the extreme severity of the stress scenario and the
exclusion of corrective management actions from the
exercise. Moreover, despite the severity of the scenario
used, in Canada’s 2013 FSAP, Canadian banks maintain
a solid ability to generate capital, which is consistent with
their past experience in times of stress. As outlined in its
report, the IMF views the resulting capital shortfall in the
FSAP stress scenario as manageable, emphasizing the
overall resilience of the Canadian banking system.
The difference between the results obtained in the
bottom-up stress test and those obtained with MFRAF
stems from the marginal impact of liquidity risk and
network spillover effects. Liquidity risk and network
effects lead to an additional 40-basis-point decline in
the aggregate CET1 ratio beyond the effect of solvency
risk. Liquidity risk explains 65 per cent of this additional
decline, and network effects account for the remaining
35 per cent.24
These results illustrate the importance of liquidity risk
and network spillover effects in times of stress: they add
almost 20 per cent to the estimated impact of this stress
scenario on banks. It is therefore important for authorities
to account for these effects when assessing the potential
impact of stress scenarios on the banking system.
Conclusion
Stress testing is an important component of the tool kit
available to authorities, including the Bank of Canada,
to assess risks to the financial system. However, it is
important to highlight that, despite recent significant
progress in the development of stress-testing models,
stress testing remains challenging because it attempts
to capture the effects of tail events.
In most stress tests, solvency risk explains a large share
of the deterioration in the capital ratios of banks during
periods of severe stress. As demonstrated by the recent
financial crisis, however, liquidity risk and network
spillover effects can generate substantial additional
losses for banks. Hence, it is important to take them into
account when assessing risks. To this end, the Bank
of Canada has developed an innovative stress-testing
model, the MacroFinancial Risk Assessment Framework
(MFRAF), which incorporates various sources of risk for
banks—solvency risk, liquidity risk and spillover effects.
Research is ongoing to improve MFRAF in two directions. First, the liquidity module could be enhanced
by developing a model to link the evolution of market
liquidity conditions with the behaviour of banks under
stress (e.g., their decision to sell liquid or illiquid assets
to meet their funding needs). Second, MFRAF should
incorporate a model of risk-weighted assets to more
accurately estimate the effects of solvency risk, liquidity
risk and network effects on bank capital levels.
22 See the report by Chouinard and Paulin in this issue on pages 53–59.
23 MFRAF was run for only the second and third years of the stress horizon
because those years were the worst period of the stress scenario in terms
of real growth and financial market conditions.
24 Network effects have a limited impact because the big banks have relatively
small interbank exposures, owing to the extensive use of collateralization
and hedging.
67
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
appendix a
The MacroFinancial Risk Assessment Framework: Model Details and Calibration
in mfRaf, the assets for each bank at the start of the
year are categorized into illiquid assets ( I ) and liquid
assets ( M ) .1 their liabilities include the stock of various liabilities that may be subject to a run in six months ( S ), other
liabilities ( L ), and their common equity tier 1 (Cet1) capital
( E ) (Figure A-1) .
in mfRaf, a bank’s liquidity characteristics are summarized by
M
I
its balance-sheet liquidity ( ), which is the ratio of the value
of liquid assets (M ) and illiquid assetsS( I ) at the expected
fire-sale discount
stress conditions to the stock of
M ( ) under
I
liabilities susceptible
to
a
run
(
S ) at the interim date:2
S
M
I
S
Figure A-1: Typical bank balance sheet at the start of the year
Illiquid assets
I
Liquid assets
M
Liabilities coming due
in 6 months
S
Other liabilities
L
Capital
Calibration
Running mfRaf requires a large amount of bank balance-sheet
data . for the international monetary fund’s 2013 financial
Stability assessment Program (fSaP), mfRaf was used as a
“hybrid” to complement the banks’ bottom-up stress test . Hence,
the data came primarily from the bottom-up stress tests and
regulatory returns (Table A-1) . the data on interbank exposures
used in the network module are from a new regulatory return
completed by major Canadian banks .
Running mfRaf also requires calibrating some elements
of the model, primarily for the liquidity-risk module . 3 the
parameters for the liquidity-risk module were calibrated to
be broadly consistent with recently introduced international
liquidity standards .4 liquid assets include cash holdings and
government and other securities that can be pledged as collateral to the liquidity facilities of central banks . illiquid assets
refer to loans to the corporate and household sectors, as well
as securities that cannot be pledged to central banks but can
be sold for cash in secondary markets (subject to large haircuts calibrated to be consistent with stressed market liquidity
conditions) . the liabilities that may be subject to a run ( S )
are obtained by aggregating the different funding instruments
and maturity profiles, taking into account their respective
degrees of stability based on their nature and maturity
(e .g ., retail deposits are more stable than wholesale funding) .
E
Source: Bank of Canada
1
for technical details on the model, see Gauthier, He and Souissi (2010) and
anand and Bédard-Pagé (2014) .
I
2 theM
term captures
the sentiments of creditors concerning the fire-sale discount
that the bank
S will suffer if it liquidates its portfolio of illiquid assets .
3 the liquidity calibration was agreed upon by Canadian authorities and the
international monetary fund . to assess the sensitivity of the results to the
liquidity calibration, a calibration that was twice as severe was also considered in
the fSaP . under this alternative liquidity calibration, the effects of liquidity risk
are more pronounced .
4 for a discussion of the international liquidity standards, see Gomes and wilkins
(2013); for details about the standards, see BCBS (2013) .
Table A-1: Data: Sources and calibration
Variables
Solvency-risk module
Liquidity-risk module
Source
EAD, PD, LGD (by economic sectors)a
Bottom-up stress test, reported by banks
Historical covariance matrix of defaults
Bank of Canada internal model
Operating income
Bottom-up stress test, reported by banks
Liquid assets ( 00)
Illiquid assets ( 0 )
0
0
0
–
M –( ) I
Fire-sale discounts
–S
Liabilities subject to a run ( 𝑆𝑆0 )
b
Network-effects module
Interbank exposures
CET1 ratio denominator
Risk-weighted assets
𝑆𝑆0
𝑆𝑆0
Regulatory data
Regulatory data
Bank of Canada calibration, based on market expertise
Regulatory data and Bank of Canada calibration based on international
liquidity standards for the inclusion of funding instruments ranked by
their stability
Regulatory data
Bottom-up stress test, reported by banks
a. EAD = exposures at default; PD = probability of default; LGD = loss given default
b. All types of interbank exposures were included, after taking into account allowable netting agreements, admissible hedging practices and the value of collateral
received. For more information, see OSFI (2013b).
Source: Bank of Canada
68
Stress Testing the Canadian Banking System: A System-Wide Approach
Bank of Canada • Financial System Review • June 2014
References
Anand, K. and G. Bédard-Pagé. 2014. “MFRAF.” Bank of
Canada Technical Report (forthcoming).
Anand, K., C. Gauthier and M. Souissi. 2014.
“Heterogeneous Beliefs and Systemic Risk.” Bank of
Canada Working Paper (forthcoming).
Basel Committee on Banking Supervision (BCBS).
2009. “Principles for Sound Stress-Testing Practices
and Supervision.” Bank for International Settlements
(May).
—.
2013. “Basel III: The Liquidity Coverage Ratio
and Liquidity Risk Monitoring Tools.” Bank for
International Settlements (January).
Brunnermeier, M. K. 2009. “Deciphering the Liquidity
and Credit Crunch 2007–2008.” Journal of Economic
Perspectives 23 (1): 77–100.
Gauthier, C., Z. He and M. Souissi. 2010.
“Understanding Systemic Risk: The Trade-Offs
Between Capital, Short-Term Funding and Liquid
Asset Holdings.” Bank of Canada Working Paper
No. 2010-29.
Gomes, T. and C. Wilkins. 2013. “The Basel III Liquidity
Standards: An Update.” Financial System Review
(June): 37–43.
International Monetary Fund (IMF). 2014. “Canada:
Financial Sector Assessment Program; Stress
Testing—Technical Note.” IMF Country Report No.
14/69.
Norges Bank. 2013. Financial Stability Report 2013.
Office of the Superintendent of Financial Institutions
(OSFI). 2009. “Stress Testing.” Guideline No. E-18
(December).
Brunnermeier, M. K. and L. H. Pedersen. 2009. “Market
Liquidity and Funding Liquidity.” Review of Financial
Studies 22 (6): 2201–38.
—.
Chouinard, E. and G. Paulin. 2014. “Making Banks Safer:
Implementing Basel III.” Financial System Review
(June): 53–59.
—.
Eisenberg, L. and T. H. Noe. 2001. “Systemic Risk in
Financial Systems.” Management Science 47 (2):
236–49.
2013a. Advisory. “Domestic Systemic
Importance and Capital Targets—DTIs. Capital.”
(March).
2013b. “Interbank and Major Exposures Return
(EB/ET)” (Revised November).
Sveriges Riksbank. 2012. Financial Stability Report
2012:2.